Oswestry Disability Index: a psychometric analysis with 1,610 patients

One-fourth of the adult US population has or will experience back pain and has undergone one of a myriad of treatments. Understanding the outcomes of these many treatments from pharmacologic to surgical, from manipulation to modality, allows for a better understanding and value-driven decision making. Patient-reported outcome measures are the current standard and include general and disease-specific measures. The Oswestry Disability Index (ODI) is the most commonly used disease-specific patient-reported outcome tool to measure functional disability related to back pain. Few studies have evaluated its psychometric properties in a large patient sample using a modern tool such as the Rasch analysis model. This study aims to identify the benefits and deficiencies of the ODI as an outcome tool for assessing patients with back pain. This study aimed to investigate the psychometric properties, performance, and applicability of the ODI in patients with back pain who visited a university-based outpatient clinic. This study used a secondary analysis—assessment of diagnostic tool on consecutive patients. The sample comprised 1,610 patients visiting an academic spine center. The ODI was the outcome measure. Detailed Rasch analysis of the ODI was performed. Standard descriptive statistics were also assessed. The ODI performed well overall. It demonstrated suboptimal unidimensionality (ie, unexplained variance after accounting for the first dimension) of 8.3%. Person reliability was good, at 0.85, and item reliability was excellent, at 1.00. The overall item fit for the ODI was good with an outfit mean square of 1.02. The ODI had a floor effect of 29.9% and ceiling effect of 3.9%. The raw score to measure correlation of the ODI was excellent, at 0.944. The ODI performed relatively well overall, with some problematic findings. It had good person and item reliability, although it did not demonstrate strong evidence of unidimensionality. The ODI has moderately poor coverage, with a very large floor effect and small ceiling effect, which could present a challenge in interpreting results of scores at the end of the spectrum.